Automated record linkage systems are increasingly used to assess the outcomes of drug therapy. Such assessments depend both on the assumption that pharmacy dispensing data is an adequate measure of drug exposure and also on the adequacy of clinical information. The accuracy of the dispensing-exposure assumption has not been rigorously tested, nor has dispensing been correlated with actual prescribing in the HMO setting. Most record linkage systems are also constrained by limited information about ambulatory conditions. This investigation will assess the usefulness of an automated record linkage system that contains pharmacy dispensing data as well as both ambulatory and inpatient clinical data for studying the indications for, outcomes of, and resource utilization associated with prescribed drug therapy. One major aim is to determine whether either prescribing or dispensing data is a sufficiently accurate measure of drug exposure to allow these to be used for investigation of several aspects of drug therapy, including determinants of prescribing, compliance, effectiveness, and adverse reactions to therapy. We will use the Harvard Community Health Plan's automated ambulatory record system to identify hypertensive patients and to determine the therapy that is prescribed. Dispensing information for these patients will be determined from automated records. We will directly assess compliance among a sample of individuals by means of an automated monitor that tracks the date and time that the medication vial is opened. The degree of compliance determined from dispensing records will be compared to compliance recorded by the medication monitors. Exposure determined by each of the three measures (prescribing, dispensing and medication monitors) will be compared to the patient's blood pressure. This information will be used to determine the outcomes of therapy, and total resource utilization associated with specific treatment regimens. It will also be used to explore reasons for failure to achieve adequate blood pressure control despite adequate dispensing. It will also allow identification of patients who are candidates for step down or discontinuation of therapy because of persistent good blood pressure control. We have chosen to study therapy of hypertension because this is one of the most common indications for medical care and because the principal outcome, blood pressure, must be determined in ambulatory records. This approach may also provide a model for investigation of drug therapy of other conditions, such as diabetes, depression and epilepsy, for which the principal outcomes are captured by automated ambulatory records rather than inpatient records.